Although Google Now, Apple Siri and Microsoft's Cortana serve tens of millions of users on mobile clients, automakers are still trying to develop a more reliable and efficient in-vehicle voice recognition system.

According to the J.D. Power 2016 American New Car Quality Survey, among the problems reported by car owners, the entertainment system and voice recognition system-related failures accounted for as high as 23%. Renee Stephens, vice president of J.D. Power's automotive quality department, said that although speech recognition has made a little progress this year, speech recognition failures are still the most common problem.

It is that only the older generation of car owners think that the voice recognition function is difficult to use. Even car owners born in the 60s and 80s generally think that the voice recognition function is the top five most difficult to use functions in the car.

Nuance Communications, a better automotive voice recognition provider in the automotive industry, has released the Dragon Drive voice assistant, indicating that voice recognition technology is constantly evolving. Arnd Weil, the company’s general manager, pointed out that last year it launched an important new 7-series sedan. After the voice software is updated, the BMW7 series sedan will become the new benchmark model for automotive voice recognition technology.

​

The new voice system allows car owners to give voice commands, which are converted into text messages through voice and sent out, and can even interrupt the operation of the system through instructions. Although speech recognition has not been fully developed, its reliability has indeed improved. Arnd Weil used phone calls as an example. The accuracy of voice calls is as high as 90% to 95%. Voice recognition is also getting better and better when using route navigation, but it is still difficult to handle commands that lack structure, such as requests to associate points of interest.

Early voice systems relied on databases, and search commands were limited to some well-known chain stores, such as "Starbucks", "Exxon Mobil Gas Station" or "Hilton Hotel". But the new on-board computer can be connected to the cloud to search for commands, provided that the owner’s commands can be "heard" clearly.

So where is the speech recognition problem? One is that there are often nouns that are difficult for machines to understand in place names, and the other is the noise of other passengers in the car.

In order to solve this problem, car manufacturers began to use two microphones to replace the original one for radio. The principle is like the two ears of a person. These two microphones can accurately point out the position of each passenger in the car that makes a sound, so as to achieve the effect of recognizing the voice of the owner. This system called "beam-forming" was first adopted by Audi, Mercedes-Benz and BMW, and more and more Volkswagen brands are now adopting it.

More powerful computer chips have been able to achieve more demanding functions. Because the system cannot always be connected to the Internet, the vehicle can provide backup voice recognition functions through the in-vehicle computer chip. This embedded system is simpler than a cloud-based speech recognition system, but it is more user-friendly.

Although the technology has improved, the voice recognition function in the car is still inferior to the performance of a 0 mobile phone. The cockpit can still be noisy, and car owners still need to pay attention to driving conditions at all times. The development of speech recognition technology will be a slow development process.